analysis of finger movement by using motion information ... · pdf filecomputer is no longer...
TRANSCRIPT
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 24
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
Abstract – Nowadays, through the advancement of science and
technology, possibility of human finger provide information into
computer is no longer question. Fingers movement and hand
motion continuously being center of research in human computer interaction (HCI) and robotic controls. Using self-develop
DataGlove, an experiment was conducted by using motion
capture System (MOCAP) equipped with five motion capture
cameras to capture human finger movements. The purpose of
this paper is to analyze voltage output from DataGlove and angle obtains from motion capture system while constructing
relationship concerning both outcomes. Polynomial equation is
considered toward the construction of fitting curve line in scatter
data. Through the end of project, differences between finger
graphs slopes will be clarify. Preliminary result of experiment exposed the newly develop DataGlove output might closely relate
into angle of finger bending movement.
Index Term— DataGlove, Finger movements, Human
Computer Interaction, Motion Capture Software (MOCAP),
Polynomial Regression
I. INTRODUCTION
Human Computer Interaction (HCI) is term used to refer an
understanding and designing of differences relationship
between people and computer [1]. According to [2], HCI
involve in various features such as command line, menus,
Papper submitted on 10 May 2013. This work is supported by the
ScienceFund Grant by the Ministry of Science, Technology and Innovation to Universiti Malaysia Perlis (01-01-15-SF0210).
M. Hazwan Ali , Advanced Intelligent Computing and Sustainability Research Group, School of Mechatronic, Universiti Malaysia Perlis
KampusPauh Putra, 02600 Arau, Perlis, MALAYSIA (e-mail: [email protected]).
KhairunizamWAN, Advanced Intelligent Computing and Sustainability Research Group, School of Mechatronic, Universiti Malaysia Perlis
KampusPauh Putra, 02600 Arau, Perlis, MALAYSIA (e-mail:
[email protected]). Nazrul H. ADNAN, Bahagian Sumber Manusia, T ingkat 17 & 18,
IbuPejabat MARA Jalan Raja Laut, 50609 Kuala Lumpur, MALAYSIA & Advanced Intelligent Computing and Sustainability Research Group,
School of Mechatronic, Universiti Malaysia Perlis KampusPauh Putra, 02600 Arau, Perlis, MALAYSIA (e-mail: [email protected]). Y.C Seah, Advanced Intelligent Computing and Sustainability Research
Group, School of Mechatronic, Universiti Malaysia Perlis KampusPauh Putra,
02600 Arau, Perlis, MALAYSIA. Juliana Aida Abu Bakar, School of Multimedia Tech & Communication
College of Arts and Sciences Universiti Utara Malaysia 06010 Sintok, Kedah, MALAYSIA (e-mail: [email protected]).
Zuradzman M Razlan , Advanced Intelligent Computing and Sustainability Research Group, School of Mechatronic, Universiti Malaysia
Perlis Kampus Pauh Putra, 02600 Arau, Perlis, MALAYSIA. (e-mail:
natural language, direct manipulation, and form fill. Through
direct manipulation, gesture of human body that contains
meaningful information [3] will be interpreted through
pointing device/graphical display [2]. DataGlove is an
example of HCI which provide information about human
finger and motion frequently used for gesture recognitions [4].
DataGlove is input device wears identical to standard glove
capable to capture physical data such as bending of finger [5].
Due to that characteristic, DataGlove often use in Virtual
Reality [6] and hand gesture application [7]-[10].
This research concern about characteristic of finger motion
achieve from preliminary experiment while modeling finger
movement into voltage and angle correlation signal. Effect of
voltage on angle result will be revising using polynomial
regression method. According to [11], polynomial regression
is form of linear regression in which the relationship between
variable x and dependent variable y is modeled as an nth order
polynomial. Consequently, assessing variable voltage and
angle using polynomial regression would estimate the
relationship among variable.
This research paper organized as subsequent; Section 2
encompasses literature review of the related researches,
problem and approach acquiring finger data. Section 3
presents the methodologies of applied procedure. Section 4
divided into 2 sections, first section states about experiment
setup whereas second section demonstrations the result of
experiments. Final section 5 expresses the conclusion over
current research.
II. LITERATURE REVIEW
The modeling of finger motions in this research is based on
GloveMAP and motion capture data. GloveMAP on the
contrary is DataGlove construct using strain gauge sensor to
measure finger flexion [12]. Assessing output voltage from
GloveMAP with proficient signal analysis and programming,
excellent result could be demonstrated. GloveMAP
achievement has been verified by numerous experiments
revolving around GloveMAP such as virtual interaction [13]
whereby the waveform produce by GloveMAP are processes
and display into virtual reality as an alternative by means of
regular Graphical User Interface (GUI) [14]. Furthermore
through GloveMAP, PCA-based finger movement and
grasping classification development [15] has been presented
successful. Angle contradictory to GloveMAP required
dissimilar procedure to obtain the coordinate and magnitude of
Analysis of Finger Movement by Using Motion Information from GloveMAP and Motion
Capture System
M.Hazwan Ali, Khairunizam WAN, Nazrul H. ADNAN, Y.C Seah, Juliana A. Abu Bakar and Zuradzman M. Razlan
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 25
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
finger flexion.
Motion capture system by Qualisys Track Manager
Software (QTM) [16] otherwise present alternative approach
in angle analysis whereby the software already equipped with
calculation to acquire both magnitude and components for
position, angle, velocity and acceleration. Motion Capture
System (MOCAP) is generally system with capability to
record the movements of human, and others motion, and then
using recorded data to animating graphics. MOCAP mostly
use in widespread commercial such as video game and movie
while as well to other area such as biomedics, biomechanic,
education and artistic. Notable usage of MOCAP is in the
study of skeletal parameter by Adam G. Kirk et al. [17]
University of California. Jonathan Maycock, et al. [18] in
2011 also manipulating MOCAP and DataGlove on robust
tracking of human hand postures for robot teaching. While on
International Joint Conference 2006, Young-Il Oh et al. [19]
display a promising research in low cost motion capture
system for PC-based immersive Virtual Environment (PIVE)
system.
III. METHODOLOGY
Fig. 1. Flowchart of GloveMAP and QTM Software
A. Flow Chart of works
Flow Chart of works shown in fig. 1 provides overview of
the proposed system. The works start with the calibration of
MOCAP before measuring finger movements . MOCAP data
are directly transferred to MOCAP’s control computer
whereas GloveMAP data transferal over microcontroller
through serial communication port. Both finger movements
data obtained from MOCAP and GloveMAP are analyzed by
using Polynomial regression method.
B. Kinematic of finger
According to [20], kinematic is the branch of conventional
mechanics that describe the motion points, bodies and systems
of bodies without consideration of the cause’s motion.
Salvador Cobos et al. [21] stated that kinematic model of
human skeleton comprised of 19 links that initiate the
corresponding human bones, and 24 degrees of freedom
(DOF) that represent the joint. That would mean four links and
five DOF for index, middle, ring and little whereas three links
and four DOF for thumb.
Fig. 2 shows detail of kinematic model for human index
finger. In this kinematic finger model, Metacarpophalangeal
joint (MCP) is modeled by two DOF universal joint label as
ϴMCP1 and ϴMCP2 however proximal Interphalangeal (PIP)
and distal Interphalangeal (DIP) similarly have one DOF label
as ϴPIP and ϴDIP individually. Although, both kinematic
modeling and MOCAP provide method to analyze angle, the
nature constrains of human hand have to be taken into account
as it refrain finger flexion in certain angle degree. Finger
motion constrain typically classified into Intrafinger constrains
and Angle range constraints . Intrafinger constrains is a
common constraint occur to same finger joint and can be
calculate by refer to (1), whereas ϴDIP is stated as finger
bending angle of Distal Phalange joint and ϴPIP is structure for
the Proximal Interphalangeal joint bending angle.
( ⁄ ) (1)
Angle range constraints otherwise a types of difficult
constraints rising to the boundary of the range concerning
finger motions of hand anatomy which follow by refer to (2).
( )
And,
(2)
START
GloveMAP start
capturing data
Motion Capture
Start capturing
Data
Computer Process Data
END
Microcontroller
Cubic Polynomial
Marker
Detection
Calibration
Data
Trigonometry
Function
Matlab
Motion Capture Data
Yes
No
No Yes
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 26
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
Fig. 2. Kinematic model human index finger
C. GloveMAP
GloveMAP shown in fig. 3 is adopted flex sensors in the
construction of DataGlove. The resistivity of the sensor
corresponds to the increasing distance between each of carbon
element inside the thin strip of flex sensor. With the change of
the resistance values, the voltage outputs can be calculated
referring to voltage the divider equation, where Vin is the flex
sensor supply voltage, R refer as the resistance of flex sensors,
whereas Vout is voltage output resulting from referring to (3).
By analyzing voltage data into Matlab, waveform as shown in
fig. 4 is obtained.
(3)
Fig. 3. GloveMAP attached with the flex sensor
Fig. 4. Voltage waveform outputted from flex sensor
D. Motion Capture
Qualisys Track Manager (QTM) is used as tracking
software due to fact that QTM is built around set of advanced
motion capture algorithms to ensuring high performance,
accuracy and low latency [16]. QTM measures the finger
movements in 3D space. Fig. 5 shows the magnitude’s
trajectories of finger movements for marker #1, #2 and #3.
Fig. 5. Magnitude of marker trajectories
E. Trigonometry Function
Trigonometric according to [22], is a branch of mathematics
that studies triangles and the relationships between the lengths
of their sides and the angle between those sides. In this
research, trigonometry function is used to model finger
movements of human. Trigonometry used in a calculation is
Pythagorean Theorem and Point-Slope Equation as written in
(4). Whereas ( , ) is a known point, m is a slope of the line
and (x, y) is any point on the line.
( ) (4)
Fig. 6 illustrates the angle ϴ°, which is determined in the
calculation. Thru expending of the equation (4), equation (5) is
produced.
ϴMCP1
ϴMCP2
ϴPIP ϴDIP
Flex sensor #1
(index finger)
LED Power
indicator
Control
Circuit
USB
Flex sensor #2
(middle finger)
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 27
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
( )
( )
( )
( ) And,
( )
( )
( )
( ) (5)
Fig. 6. Angle, Ɵ
Based on the 3 coordinates x, y and z, the correspondence Ɵ
is calculated. Fig. 7(a) shows the magnitude of marker
trajectories for each marker #1, #2 and #3. Fig. 7 (b) shows the
correspondence angles Ɵ calculated by the system.
Fig. 7. (a) Magnitude of marker trajectories for marker #1, #2 and #3; (b)
Angle, Ɵ
F. Polynomial Regression
Polynomial regression method has been used attentively to
nonlinear functions for modeling real-life phenomena and
usually used in mathematical model to expecting dependent
variable y on independent variable x. First degree of regression
analysis (nonconstant linear function) can be used in
constructing best fit straight line in scatter plot data. Second
degree polynomial is a quadratic polynomial, with better data
fit than first degree polynomial. A cubic function is a
polynomial with degree of 3 and has form as refer to (6) [23].
( ) (6)
Cubic polynomial usually provides superior data fit than first
and second order while ensuring high coefficient of
determination on scatter plots. Fig. 8 shows the result of
employing cubic polynomial into the Ɵ signal in fig. 7(b). The
cubic polynomial curve display great comparable with original
signal while expecting the subsequent signal sequence.
Fig. 8. Cubic polynomial plotted graph
IV. EXPERIMENT
Experiments were carried out in MOCAP environment.
Both GloveMAP and QTM Software were simultaneously run
to read finger movements .
A. Experiment Setup
In the experiment, 3 markers were attached to GloveMAP as
shoen in fig. 11. The markers positions were at distal
phalanges, proximal phalanges and metacarpals of the index
finger. The cameras used in the experiment were built on Oqus
100 with average residual of 0.3 to 0.9 mm at 3.7 m at
distance. The experiments were conducted by doing various
movements of index finger. Each experiment was run in 2 s.
ϴ
Marker#1
Marker#2
Marker#3
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 28
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
Fig. 9. MOCAP environment with 5 Oqus 100 cameras
Fig. 10. Anatomy of hand [24]
Fig. 11. GloveMAP and the location of markers placement
B. Experiment Result
For the experimental result, all data obtained was analyzed
by using cubic polynomial. Angle and voltage were plotted
into same graph which voltage performs as dependent variable
y whereas angle as independent variable x. Although both
systems run for 2 s, data sampling rate for GloveMAP and
QTM was difference. The sampling rate of QTM was 100 fps
whereas GloveMAP contain 20 fps. Fig.12 (a) shows the
voltage signal of GloveMAP and fig.12 (b) shows angle signal
of QTM Software. Fig. 12 (c) shows the resampled signal of
angle based on the obtained voltage data.
Fig. 12. Signals (a) Voltage signal (b) Angle signal (c) resampling angle
signal
Finger movements have the directions, which were bending
and straighten movements. Fig. 13(a) shows signal gradient
decline as angle increase indicating the finger was bent
meanwhile fig.13 (b) shows slope rise with angle decreasing
indicate that finger was straighten. Fig.13 (c) and (d) show the
results after employing polynomial regression of finger
movements data. The value of correlation coefficient
indicates linearity relationship between the correlation
Marker#3
Marker#2
Marker#1
Flexible Bend Sensor
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 29
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
coordinate points . Fig.13 (c) and (d) have r2 values of 0.99 and
0.991, respectively. The strong correlation or relationship is
defined that has a value ranging between 0.85 to 1 [25].
Norm of residual value was also observed. Norm of residual
is defined as the difference between observe value with the
estimate function value of the unobservable statically error
[26]. It also may refer as measure of the deviation between
the correlation and data. A lower norm of residual value
implies a better fit of regression to the observe data. Norm of
residual for fig. 13 (a) and fig. 13 (b) is 0.056058 and 0.04691,
respectively. The values show that a small degree of error
when finger bending and straighten. Moreover, the patterns of
dot slopes designate that relationship between GloveMAP
output voltage with angle are may possibly perpendicular to
each other.
Fig. 13. Graph of finger flexion with polynomial (a) finger bend (b) finger
straighten (c) finger bend polynomial (d) finger straighten polynomial
V. DISCUSSIONS
Based on the results, the correlation between voltage signals
outputted from the DataGlove and the angle calculated from
motion data acquired from MOCAP is established. The scope
of research works is to find the correlation between the
voltages outputted from the proposed DataGlove with the
angle of finger’s movement. The correlation could be used in
future experiments for the acquisition of finger movement’s
data of the proposed DataGlove. However, human fingers
have many degrees of freedom, and in the future various
experiments need to be done as a further analysis to evaluate
the performance of the proposed DataGlove.
VI. CONCLUSIONS
The research works proposed the analysis of finger
movements by using polynomial regression approach. The
DataGlove called GloveMAP is used in the experiment.
GloveMAP is a low cost DataGlove developed by our research
group. The output signal from GloveMAP is a voltage signal.
MOCAP system is used to measure the bending angle of the
finger. The analysis is done to correlate the output voltage and
the bending angle of finger by using polynomial regression
approach. The experimental results show that the equation that
represents the correlation between output voltage and angle
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 30
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
could be produced. Furthermore, finger movements include
bending and straighten. The experimental results show that
bending and straighten movements have a similar
characteristic. In the future, the results will be used to acquire
various signals of grasping activities. The classifier will be
employed to train computer knows how to grab various
objects based on shapes and patterns.
ACKNOWLEDGMENT
Special thanks to all members of UNIMAP Advanced
Intelligent Computing and Sustainability Research Group and
School Of Mechatronics Engineering, Universiti Malaysia
Perlis (UNIMAP) for providing the research equipment’s and
internal foundations. This work is supported by the
ScienceFund Grant by the Ministry of Science, Technology
and Innovation to Universiti Malaysia Perlis (01-01-15-
SF0210).
REFERENCES [1] Richard Harper et al., “Being Human: Human-Computer
Interaction in the year 2020,” Microsoft Research Ltd, 2008, pp.
43.
[2] Sean Bechhofer, “Human Computer Interaction,” Lecture note
University of Manchester, 2010, pp. 13-16.
[3] Mark Billinghurst , “Gesture Based Interaction,” Lecture note
Chapter 14, Aug.24, 2011, pp.1.
[4] David L. Quam, “Gesture Recognition with a DataGlove,”
National Conference on Aerospace and Electronics (NAECON 1990), IEEE, 1990, PP. 775-760.
[5] Information on http://en.wikipedia.org/wiki/Wired_glove
[6] S. Sayeed, N. S. Kamel, and R. Besar, “Virtual Reality Based
Dynamic Signature Verification Using Data glove,” International
Conference on Intelligent and Advanced Systems 2007 (ICIAS
2007), Nov.25-28, 2007, pp. 1260-1264.
[7] M. Ishikaws and H. Matsumnra, “Recognition of a Hand-Gesture
Based on Self-organization Using a DataGlove,” 6th International Conference on Neural Information Processing (ICONIP '99), vol.
2, 1999, pp. 739-745.
[8] S. Saengsri, V. Niennattrakul, and C. A. Ratanamahatana, “ TFRS:
Thai Finger-Spelling Sign La Recognition System,” 2012 Second
International Conference on Digital Information and
Communication Technology and it’s Applications (DICTAP),
2012, pp. 457-462.
[9] Wu jiangqin et al., “A Simple Sign Language Recognition System
Based on DataGlove,” 1998 Fourth International Conference on Signal Processing Proceedings (ICSP '98), vol. 2, 1998, pp. 1257-
1260.
[10] T. T . Swee et al., “Wireless Data Gloves Malay Sign Language
Recognition System,” Information, Communications & Signal
Processing (ICICS 2007), Dec. 10-13, 2007, pp. 1-4.
[11] Information on http://en.wikipedia.org/wiki/Polynomial_regression
[12] Nazrul H. ADNAN et al., “The Development of a Low Cost Data
Glove by Using Flexible Bend Sensor for Resistive Interfaces,” The 2nd International Malaysia-Ireland Joint Symposium on
Engineering, Science and Business 2012 (IMiEJS2012) , 2012, pp.
579-587.
[13] Nazrul H. ADNAN et al., “Measurement of the Flexible Bending
Force of the Index and Middle Fingers for Virtual Interaction,”
International Symposium on Robotics and Intelligent Sensors 2012
(IRIS 2012), Procedia Engineering 41 , 2012, pp. 388-394.
[14] Y. A. A. Refaat , and A. A. Ahmed, “Introduction to Graphical User Interface (GUI) MATLAB 6.5,” IEEE UAEU Student Branch
UAE University, pp. 2-35.
[15] Nazrul H. ADNAN et al, “PCA-based Finger Movement and
Grasping Classification using Data Glove “Glove MAP”,”
International Journal of Innovative Technology and Exploring
Engineering (IJITEE), 3rd
ed, vol. 2, Feb. 2013, pp. 66-71.
[16] Information on http://www.qualisys.com/products/ Software/qtm/
[17] A. G. Kirk, J. F. O’Brien, D. A. Forsyth, “Skeletal Parameter
Estimation from Optical Motion Capture Data,” IEEE Computer
Society Conference on Computer Vision and Pattern Recognition
(CVPR 2005), vol. 2, June 20-25, 2005. [18] J. Maycock, J. Steffen, H. Ritter, “Robust Dataglove Mapping for
Recording Human Hand Postures,” 4th International Conference
(ICIRA 201)1, Aachen, Germany, Proceedings, Part II, Dec. 6-8,
2011, pp. 34-45.
[19] Young-Il Oh, Kyoung-Hwan Jo, and Jihong Lee, “Low Cost
Motion Capture System for PC-based Immersive Virtual
Environment (PIVE) System,” International Joint Conference
2006 (SICE-ICASE), Bexco, Busan, Korea, Oct. 18-21, 2006, pp.
3527-3530. [20] Information on http://physics.tutorvista.com/motion/ kinematics-
equations.html
[21] S. Cobos, M. Ferre, R. Aracil, J. Ortego and M. A. Sanchez-Uran,
(2010), “Simplified Human Hand Models for Manipulation
Tasks,” Cutting Edge Robotics 2010 Chapter 10, pp. 156,
Available: http://www.intechopen.com/books/cutting-edge-
robotics-2010/simplified-human-hand-models-for-manipulation-
tasks [22] Information on http://en.wikipedia.org/wiki/ Trigonometry
[23] Information on http://dufu.math.ncu.edu.tw/calculus/
calculus_pre/node7.html
[24] Information on http://www.yalemedicalgroup.org/stw/
Page.asp?Page ID=STW023547
[25] [25]Information on http://www.isixsigma.com/tools-
templates/graphical-analysis-charts/understanding-scatter-
diagrams-and-correlation-analysis.html [26] [26]Information on http://en.wikipedia.org/wiki/Errors_and_
[27] residuals_in_statistics.
M.Hazwan Ali received his Bachelor Engineering
degree in Mechatronic Engineering from University
Malaysia Perlis in 2012. He is currently a MSc
student at University Malaysia Perlis. His research
interest is in Robotics, Human-Computer
Interaction (HCI), Virtual Reality and Artificial Intelligence.
Khairunizam WAN received his B. Eng. degree in
Electrical & Electronic Eng. from Yamaguchi
University and Ph.D. in Mechatronic Eng. from
Kagawa University, in 1999 and 2009 respectively.
He is currently a Senior Lecturer at School Of
Mechatronic Engineering, University Malaysia
Perlis. He is member of Board of Engineer and Institute of Engineer, Malaysia. His research
interest is in Human-Computer Interaction (HCI),
Intelligent Transportation System, Artificial
Intelligence and Robotics.
Nazrul H. ADNAN received his Bachelor
Engineering (Hons) in Power Electrical from
Universiti Teknologi MARA (UiTM) and Master Engineering in Advanced Manufacturing
Technology from Universiti Teknologi Malaysia
(UTM) since 2004 and 2010 respectively. After
graduated in Bachelor Engineering he joined Majlis
Amanah Rakyat (MARA) as Teaching Engineer
where he worked as a lecturer to Mechatronics,
Electronics and Mechanical Diploma students. He
was currently a PhD student in Universiti Malaysia Perlis. His research interest is in Human-Computer Interaction (HCI), Product
Design, Artificial Intelligence, and Machine Design.
International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:03 31
136803-5757-IJMME-IJENS © June 2013 IJENS I J E N S
Y. C. Seah is currently a Bachelor Degree student in
Mechatronic Engineering at University Malaysia
Perlis. His research interest is in Robotic, Artificial
Intelligence and Mechatronic System,
Juliana A. AbuBakar currently lectures virtual
reality and multimedia technology courses at
University Utara Malaysia (UUM). She received
B.Eng. degree in Electronic Engineering from
University of Leeds, UK in 1999 and MSc. degree in
Information Technology from UUM in 2003. She was awarded Ph.D from International Islamic University
Malaysia in 2012 where her Ph.D thesis covers a
complete cycle of design, development, and user
evaluation of a virtual reality application for
architectural heritage learning. She is passionate in virtual reality research and
development projects since her first involvement in the academic world and
has secured several national grants and published many articles in the area.
Zuradzman M. Razlan received his Bachelor of
Mechanical Engineering from Yamagata University,
Japan from Apr 1993-Mar 1997 and Ph.D. in
Engineering from Mie University, Japan. He is
currently Senior Lecturer at School Of Mechanical
Engineering, University Malaysia Perlis and
experience in his field almost 11 years in R&D
Design Engineering. His research interest in Energy, Thermo-Fluid, Two Phase Flow, Air Flow System,
Heat-Pump and refrigeration system.